The triple isotope composition (δ17O and δ13O) of dissolved O2 in the ocean and in ice cores was recently used to assess the primary productivity over broad spatial and temporal scales. However, assessment of the productivity with the aid of this method must rely on accurate measurements of the 17O/16O versus 18O/16O relationship in each of the main oxygen-producing and -consuming reactions. Data obtained here showed that cleavage of water in photosystem II did not fractionate oxygen isotopes; the δ18O and δ17O of the O2 evolved were essentially identical to those of the substrate water. The fractionation slopes for the oxygenase reaction of Rubisco and respiration were identical (0.518 + 0.001) and that of glycolate oxidation was 0.503 + 0.002. There was a considerable difference in the slopes of O2 photoreduction (the Mehler reaction) in the cyanobacterium Synechacystis sp. strain PCC 6803 (0.497 + 0.004) and that of pea (Pisum sativum) thylakoids (0.526 + 0.001). These values provided clear and independent evidence that the mechanism of O 2 photoreduction differs between higher plants and cyanobacteria. We used our method to assess the magnitude of O2 photoreduction in cyanobacterial cells maintained under conditions where photorespiration was negligible. It was found that electron flow to O2 can be as high as 40% that leaving photosystem II, whereas respiratory activity in the light is only 6%. The implications of our findings to the evaluation of specific O 2-producing or -consuming reactions, in vivo, are discussed.